The Effectiveness of Oxygen Delivery and Reliability of Carbon Dioxide Waveforms: A Crossover Comparison of 4 Nasal Cannulae : Anesthesia & Analgesia

Secondary Logo

Journal Logo

Technology, Computing, and Simulation: Research Report

The Effectiveness of Oxygen Delivery and Reliability of Carbon Dioxide Waveforms

A Crossover Comparison of 4 Nasal Cannulae

Ebert, Thomas J. MD, PhD; Novalija, Jutta MD, PhD; Uhrich, Toni D. MS; Barney, Jill A. MS

Author Information
doi: 10.1213/ANE.0000000000000537
  • Free

In July of 2011, the American Society of Anesthesiologists (ASA) amended its standard for basic anesthetic monitoring to include exhaled, end-tidal carbon dioxide (ETCO2) monitoring during moderate and deep sedation. This has increased the need for an accurate and reliable means of noninvasively monitoring ETCO2 from respiratory devices. The most convenient, and perhaps least expensive method to monitor ETCO2 is via the nasal cannula (NC).

The first CO2 sampling NCs were introduced into clinical practice in the late 1980s to early 1990s.1–5 Since then, a number of manufacturers have entered the market, providing a variety of NCs, each with unique design features. All NCs sample ETCO2 through the nasal prongs. Some NCs have separate prongs in which O2 is delivered through one prong and CO2 is sampled via the other prong of the cannula. Some have bifurcated nasal prongs that deliver O2 and sample CO2 within each nasal prong. Finally, some provide a “blow-by” O2 cloud outside the nares via small holes in the hub of the NC, while the 2 nasal prongs are used to sample CO2. Each demonstrates efficacy for their purpose of O2 delivery. However, the designs have not been rigorously evaluated for their accuracy of ETCO2 measurements across higher flow rates of O2 delivery, which can dilute the exhaled CO2, leading to unreliable detection of respiratory function. The different designs might also affect the ability to increase the partial pressure of O2 when increasing the flow rate of O2. Thus, in the present study, we evaluated 4 different NC designs from 4 manufacturers and determined the reliability of each one for sampling ETCO2 and in delivering O2 with increasing flow rates.


After IRB approval, written informed consent was obtained from 45 study volunteers who were nonsmokers, 18 to 35 years old, and without cardiopulmonary disease. Four NCs were evaluated (Salter Labs® #4707, Arvin, CA; Oridion Microstream O2/CO2 #006912, Bedford, MA; Hudson Softec Bi-Flo #1845, Research Triangle Park, NC; Medline Mac-Safe #369-#, Mundelein, IL) (Fig. 1). The Oridion and Medline designs deliver O2 from a “cloud” via what is referred to as blow-by delivery. The Oridion uses multiple vents (multivented) and the Medline uses dual vents (dual-vented) to produce the O2 cloud. The Hudson NC has bifurcated nasal prongs (bifurcated) to deliver O2 and sample CO2 from both nares. The Salter NC uses separate nasal prongs (separate), one to deliver O2 and the other to sample CO2. These 4 NCs were compared for effectiveness in delivering O2 and sampling of ETCO2 in a randomized, prospective, crossover study design.

Figure 1:
Close-up photographs of the 4 nasal cannula (NC) designs; arrows indicate direction of air flow (O2 delivery and CO2 sensing). Clockwise from the top left: Bifurcated cannula (Hudson Softec Bi-Flo® #1845); O2 blow-by via dual vents (Medline Mac-Safe HCW 4561); O2 blow-by via multi vents (Oridion MicrostreamFilterLine #006912); and separate O2 delivery/CO2 sampling nasal prongs (Salter Labs® #4707). Multiple vents for O2 delivery can be seen in the Oridion NC, whereas dual vents can be seen in the Medline NC; both NCs supply an O2 cloud with the blow-by technique.


In a subset of 11 volunteers, a 20-gauge, radial artery catheter was inserted under aseptic conditions and used to obtain arterial blood gas measurements of the partial pressures of oxygen and carbon dioxide (PaO2, PaCO2) during baseline and 2 and 4 Lpm fresh gas flows (FGFs). Blood gases were determined using an i-Stat® System (Abbott Point of Care Inc., Princeton, NJ), with the calibrations recommended by the manufacturer. In all volunteers, electrocardiogram electrodes were placed to monitor heart rate. Pharyngeal O2 samples were obtained via a Hauge Airway (Sharn Anesthesia Products, Tampa, FL) positioned into the mouth such that the distal tip of the device was sitting in the posterior pharynx. ETCO2 was obtained by sidestream sampling from the nasal prong(s) at 300 mL/min, and infrared spectroscopy was used to determine gas concentrations (Oxigraph O2 Cap Analyzer, Mountain View, CA).

Study Design and Protocol

Intervention assignments (NC placements) for each subject were administered using a fixed, randomized block design. Volunteers served as their own control in a 4-period (room air [RA], 2, 4, and 6 Lpm FGF), crossover study design.

Subjects were instructed to breathe normally during a protocol of step increases in FGF with each NC. The research assistant positioned the first NC, and nasal breathing was confirmed by the presence of an acceptable CO2 tracing. The Hauge Airway was placed and the subject breathed RA while ETCO2 and posterior pharyngeal O2 data were collected continuously throughout the period. In the 11 volunteers with an arterial catheter, PaO2 and PaCO2 also were measured. The O2 FGF then was set to 2 Lpm, and again, data were collected after achieving stable recordings (Fig. 2). This procedure was repeated at O2 FGF of 4 and 6 Lpm. After data collection at 6 Lpm, the first NC was removed and the second NC was placed. Procedures were then repeated with RA, 2, 4, and 6 Lpm O2. These same step-up procedures were repeated with the third and fourth NCs. The Hauge Airway was briefly removed if the subject needed to swallow or was simply uncomfortable between NC placements. The additional time for sampling and processing of arterial blood samples from the subset of 11 volunteers caused us to omit blood sampling during the 6 Lpm study period, resulting in blood gas data at RA, 2, and 4 Lpm FGF.

Figure 2:
Representative tracings from 2 nasal cannulae during 2 Lpm O2 FGF. Pharyngeal O2 concentration is represented with the horizontal line running through the CO2 waveforms. ETCO2 analysis was initiated when pharyngeal O2 had stabilized during each level of O2 FGF. The top tracing shows normal CO2 waveforms during the increase in FGF to achieve a steady state at 2 Lpm O2 FGF. The bottom tracing demonstrates a damped ETCO2 tracing during a similar period. Because of the damped waveform, the calculated ETCO2 was notably reduced.

Data Analyses

Analysis of electronic files was performed in a blinded fashion. Raw data were securely transferred from the research lab to an independent statistician throughout the study. The independent statistician then masked all data files as to subject, NC, and flow rate information. Moreover, the statistician randomized the sequence in which the data files were to be reviewed. These deidentified files were then securely returned to the research lab en masse for retrospective analysis by the investigators after completion of data collection. The statistician confirmed via a training exercise that research personnel who were to perform the analysis provided accurate, reproducible, and reliable analysis of the files. Files were carefully examined and when an extended period of normal breathing with stable pharyngeal O2 was found, at least 1 minute of continuous pharyngeal O2 and ETCO2 data were averaged for each condition (RA, 2, 4, and 6 Lpm). The sampled ETCO2 and pharyngeal O2 waveforms are shown in the shaded area in the right side of Figure 2. These 2 examples were selected to demonstrate the variation observed in ETCO2 waveforms from 2 different NCs as O2 is delivered at 2 Lpm FGF.

Prestudy power analysis and sample size calculations were performed based on the results from an internal 10-volunteer study from Salter Labs® (#C002 entitled “ETCO2 Cannula Clinical Study Report”). It was determined that a sample size of 4 in each cannula group would have 80% power to detect a difference in overall ETCO2 means of 0.90, the difference between the Salter 4707 Cannula Group of 44.2, and the Oridion Cannula Group mean of 43.3, assuming that the common standard deviation is 0.3 using a 2-group matched pairs t test with a 0.05 2-sided significance level. Although the results from this previous study indicate that only 4 subjects in each cannula group would be needed to detect a statistically significant difference in ETCO2 means, the Sponsor requested an enrollment of 45 subjects to assure the robustness of the findings.


Descriptive statistics (mean ± SD) and exploratory data analysis were used to generate measures of central tendency and to perform data distribution diagnostics. The majority of the statistical analysis entailed using the Tukey-Kramer honest significant difference test for multiple mean comparisons. All P values and confidence intervals for pairwise mean differences were Tukey corrected. Analysis of means was used to compare variances across groups. These methods test whether the NC group standard deviations were statistically different from the root mean square error. Moreover, multiple linear regression methods were used to explore the relationship between ETCO2 and pharyngeal O2 after adjusting for cannula type and flow rate. Analyses were performed with SAS Analytics Pro, Version 9.3 and JMP Statistical Software, Version 11 (SAS Institute Inc., Cary, NC). A P < 0.05 defined statistical significance.


Forty-five volunteers participated: 28 females and 17 males, mean age 24 ± 4 years, height 172 ± 10 cm, weight 68 ± 13 kg (mean ± SD). Of those 45, 11 volunteers had an invasive (arterial catheterization) procedure for the measurement of PaO2 and PaCO2: 8 females and 3 males, mean age 24 ± 5 years, height 169 ± 8 cm, weight 63 ± 9 kg. No volunteer had craniofacial abnormalities or nasal congestion. One volunteer developed mild nasal secretions (a runny nose) requiring a tissue during the study; those data were included in the analysis because the responses were appropriate. The remaining 44 volunteers tolerated the NCs well. Most volunteers removed and reinserted the Hauge Airway when the level of O2 FGF was being ramped to the next level. This was generally for a period of approximately 5 seconds to swallow. The Hauge Airway also was removed whenever the NC was changed.

Figure 3 displays the observed ETCO2 means and standard deviations at all O2 flow rates. Using ANOM methods to compare group variances, it was found that the upper limit of the ETCO2 standard deviation for the bifurcated NC exceeded the upper limit of the standard deviation for all the other NCs (all P < 0.0001) at all flow rates. There were no significant differences in ETCO2 between the separate or multi-vented and dual-vented blow-by NCs. Mean PaCO2 (arterial blood gas, n = 11) did not differ significantly based on NC type at baseline or at the flow rates tested (averaging 39–40 mm Hg across all NCs and all experimental conditions).

Figure 3:
ETCO2 from 4 nasal cannulae (NCs) at steady state for each experimental setting (mean ± SD). The separate NC samples ETCO2 via one prong and delivers O2 via the other. In the bifurcated NC, each nasal prong is split internally with half sensing ETCO2 and half delivering O2. *Significantly higher than the bifurcated NC (P < 0.05).

Figure 4 and Table 1 display the observed pharyngeal O2 means and standard deviations at all O2 flow rates; Table 1 also includes the confidence intervals. Tukey-corrected pairwise comparison of means yielded significant differences with higher pharyngeal O2 percent with the separate (Salter) NC at 2 (95% CI: 0.8%, 3.7%), 4 (95% CI: 4.8%, 10.0%), and 6 (95% CI: 9.9%, 17.3%) Lpm compared to the multi-vented, P = 0.0005, <0.0001, and <0.0001, respectively, and at 2 (95% CI: 0.4%, 3.3%), 4 (95% CI: 2.6%, 7.8%), and 6 (95% CI: 5.2%, 12.6%) Lpm compared to dual-vented, P = 0.0055, <0.0001, <0.0001, respectively. The bifurcated NC had a significantly higher pharyngeal O2 percent than the multi-vented blow-by NC at 2 Lpm (95% CI: 0.0%, 2.9%, P = 0.0498), and higher than both the multi-vented and dual-vented blow-by NCs at 4 and 6 Lpm: multi-vented (95% CI: 3.7%, 8.9%, P < 0.0001; and 95% CI: 7.4%, 14.8%, <0.0001, respectively) and dual-vented (95% CI: 1.5%, 6.8%, P = 0.0004; and 95% CI: 2.7%, 10.1%, <0.0001, respectively). At 6 Lpm, the mean difference in pharyngeal O2 percent was significantly higher for the dual-vented blow-by NC than the multi-vented blow-by NC (95% CI: 1.0%, 8.4%, P = 0.0066). Table 2 summarizes the pairwise comparisons of mean pharyngeal O2% for all pairs of NCs at 2, 4, and 6 Lpm flow rate using Tukey-Kramer honest significant difference.

Figure 4:
Pharyngeal O2 during room air, 2, 4, and 6 Lpm FGF. The separate NC consistently provided a higher pharyngeal O2 than the blow-by NCs at 2, 4, and 6 Lpm O2 FGF (*P < 0.05); the bifurcated NC provided a higher pharyngeal O2 than the blow-by multi vent NC at 2 Lpm O2 FGF (†P < 0.05) and than both blow-by NCs at 4 and 6 Lpm O2 FGF (*P < 0.05).
Table 1:
Pharyngeal O2 (%) Measured at Baseline (Room Air), and 2, 4, and 6 Lpm Oxygen Flow Rates
Table 2:
Pairwise Comparisons of All Mean Pharyngeal O2% Using Tukey-Kramer Honest Significant Difference: Ordered Differences Report for 2, 4, and 6 Lpm Flow Rates

Mean PaO2 (arterial blood gas, n = 11) with the separate and bifurcated nasal prong NCs was statistically higher than the PaO2 achieved with the multi-vented blow-by NC at 4 Lpm O2 (Fig. 5), P = 0.004 and P = 0.03, respectively, Table 3.

Figure 5:
PaO2 from arterial blood gas analysis during room air and at 2 and 4 Lpm FGF. At 4 Lpm FGF, the 2 NCs that deliver O2 via the nasal prongs (bifurcated and separate) resulted in higher PaO2 (*P < 0.05), than the 2 blow-by NCs that deliver O2 via a cloud from vents outside the nares (multi vents and dual vents).
Table 3:
Pao2 (mm Hg) Measured at Baseline (Room Air), and 2 and 4 Lpm Flow Rates

When exploring the relationship between ETCO2 and pharyngeal O2 (from the Hauge airway in the posterior pharynx in all 45 subjects) using a partial F-test, we found a statistically significant association between ETCO2 and average pharyngeal O2 without cannula type and flow rate in the model (P < 0.0001). However, after adjusting for cannula type and flow rate, the association between ETCO2 and average pharyngeal O2 was not statistically significant (P = 0.074).

When cannula type was added to the model with average pharyngeal O2 already in the model, the association between ETCO2 and cannula type was statistically significant (P < 0.0001). In addition, a statistically significant association was found between ETCO2 and flow rate without cannula type and average pharyngeal O2 in the model (P = 0.0002). However, the association between ETCO2 and flow rate was not statistically significant (P = 0.410) after adjusting for average pharyngeal O2 and cannula type.

The analyses of noninvasive data (excluding blood gas measurements) indicate that average pharyngeal O2, cannula type, and flow rate are significant independent predictors of the variation in ETCO2. After adjusting for the other variables in the model, cannula type remained a strong predictor of ETCO2 while the associations with the other predictor variables (average pharyngeal O2 and flow rate) were not statistically significant at the 0.05 level.


NC selection is rarely driven by scientific evaluation of efficacy6; however, the present study has found noteworthy limitations of NC accuracy linked to the design of the NC. The bifurcated NC design, which delivers O2 and samples CO2 from the same nasal prong in each nares, failed to allow accurate measurements of ETCO2 at 2, 4, and 6 Lpm FGF. The multivented NC using the blow-by O2 cloud did not achieve the level of oxygenation noted when O2 was delivered into either the bifurcated NC or separate nasal prongs NCs at higher FGFs, based on both blood gas analyses and pharyngeal O2 concentrations.

The ASA has recently emphasized the importance of monitoring the adequacy of ventilation in sedation cases. The House of Delegates approved a change in the ASA Standards for Basic Anesthetic Monitoring that came into effect July 1, 2011. The new standards read as follows: “… During moderate or deep sedation, the adequacy of ventilation shall be evaluated by continual observation of qualitative clinical signs and monitoring for the presence of exhaled carbon dioxide unless precluded or invalidated by the nature of the patient, procedure or equipment.”

The added verbiage specific to sensing exhaled CO2 emphasizes the importance of monitoring ventilation in addition to O2 saturation for providing the safe conduct of moderate to deep sedation.

NC monitoring has 2 features that are interrelated. When positioned well and not diluted by RA during hypoventilation or by high flow rates of O2, the measured ETCO2 reflects PaCO2 within a few mm Hg. The accuracy of the measurement is influenced by both physiologic and anatomic dead space and shunt. Of note, all 4 NC designs studied sampled ETCO2.

For accurate ETCO2 monitoring, the bifurcated nasal prong design with both O2 delivery and CO2 sampling within the same nasal prongs appears to be a flawed design. The delivery of just 2 Lpm O2 significantly attenuated the ETCO2 detection, consistent with the report by Roth et al.7 in 1994. The blow-by multivented and dual-vented NCs that use both nasal prongs to sample ETCO2 did not result in any limitations with ETCO2 monitoring at the higher FGFs of 4 and 6 Lpm. Although one study noted that at 6 Lpm FGF, the O2 cloud system is capable of diluting ETCO2 measurements,8 the data from this study do not support the earlier observation of an attenuation of the ETCO2 data with this design. The second feature of NC monitoring is to detect apnea. Sidestream capnography has been validated as a reliable apnea monitor in patients.8

For delivery of O2, the multivented NC blow-by design that uses a cloud of O2 from holes at the base of the nasal prongs resulted in lower delivery of O2 at higher FGFs (Figs. 4 and 5). Obviously, there are multiple factors affecting the true O2 concentration in a patient receiving O2 in the perioperative setting; specifically, mouth breathing, nasal anatomy, nasal obstruction, and/or the rate and depth of respirations will change the actual O2 delivery to the patient. The separate and bifurcated nasal prongs NC designs where O2 was delivered via the nasal prongs were found to be significantly more efficient in O2 delivery compared to the blow-by NCs.

One NC design was effective in both the delivery of O2 and the sampling of CO2. The NC with separate nasal prongs, with one nasal prong to deliver O2 and the other prong to sample CO2, did not have limitations with ETCO2 sampling at higher FGFs. Additionally, of the 4 NCs studied, the separate nasal prongs design NC consistently achieved higher pharyngeal O2 percentages than the vented NCs at all FGFs tested. These combined attributes are arguably highly desirable and beneficial during moderate sedation, in the postanesthesia care unit setting, and/or when needing to deliver higher FGF to maintain saturation in disease states such as chronic obstructive pulmonary disease or morbid obesity.


This study utilized unsedated volunteers and may not apply directly to patients with cardiorespiratory disease or abnormal airways, although the performance of one cannula to effectively deliver O2 and to monitor ETCO2 appears due to a design feature that should apply across populations. However, because the performance of the Salter cannula is design related, its benefit might become a deficiency in patients with a deviated nasal septum or unilateral nasal obstruction. One of the known limitations of sidestream capnography is the potential inaccuracy of ETCO2 measurements during hypoventilation, low tidal volume breathing, and mouth breathing. In these settings, sidestream aspiration of exhaled air from the nasal prongs can become diluted with ambient air. This would lead to an underestimation of ETCO2 due to the dilution. However, this study was not designed to determine performance in these settings. Rather, we sought to determine the effect of NC design on ETCO2 measurements when the rate of O2 delivery was increased during normal ventilation.

Oral guides attached to the NC are becoming more commonplace to help sample exhaled CO2 in patients who are mouth breathers. One noteworthy investigation using an O2 flow of 4 Lpm demonstrated that the addition of an oral guide to aid in the sensing of exhalation improved the reliability of ETCO2 measurements, particularly in the obese patient with obstructive sleep apnea.9 NCs with oral guides were not evaluated in the current study.


NC design differences can influence their ability to deliver O2 and their ability to accurately sample ETCO2 at higher FGFs. A design where O2 is delivered through one nasal prong and CO2 is detected from the other prong was found to be most effective and accurate for these purposes.


Name: Thomas J. Ebert, MD, PhD.

Contribution: Thomas J. Ebert contributed to study design, conduct of study, data analysis, and manuscript preparation.

Attestation: Thomas J. Ebert approved the final manuscript.

Conflicts of Interest: The principal investigator of the study was funded by Salter Labs, manufacturer of one of the nasal cannulae studied.

Name: Jutta Novalija, MD, PhD.

Contribution: This author contributed to study design, conduct of study, and data analysis.

Attestation: Jutta Novalija approved the final manuscript.

Conflicts of Interest: The author declares no conflicts of interest.

Name: Toni D. Uhrich, MS.

Contribution: This author contributed to study design, conduct of study, and manuscript preparation.

Attestation: Toni D. Uhrich approved the final manuscript.

Conflicts of Interest: The author declares no conflicts of interest.

Name: Jill A. Barney, MS.

Contribution: This author contributed to study design and manuscript preparation.

Attestation: Jill A. Barney approved the final manuscript.

Conflicts of Interest: The author declares no conflicts of interest.

This manuscript was handled by: Maxime Cannesson, MD, PhD.


The authors wish to acknowledge the statistical services provided by Discovery Statistics, 116 Calle Patricia #A, San Clemente, CA 92672. This material is the result of work supported with resources and the use of facilities at the Clement J. Zablocki Veterans Affairs Medical Center, Milwaukee, WI.


1. Bowe EA, Boysen PG, Broome JA, Klein EF Jr. Accurate determination of end-tidal carbon dioxide during administration of oxygen by nasal cannulae. J Clin Monit. 1989;5:105–10
2. Roy J, McNulty SE, Torjman MC. An improved nasal prong apparatus for end-tidal carbon dioxide monitoring in awake, sedated patients. J Clin Monit. 1991;7:249–52
3. Liu SY, Lee TS, Bongard F. Accuracy of capnography in nonintubated surgical patients. Chest. 1992;102:1512–5
4. Dunphy JA. Accuracy of expired carbon dioxide partial pressure sampled from a nasal cannula. II. Anesthesiology. 1988;68:960–1
5. Urmey WF. Accuracy of expired carbon dioxide partial pressure sampled from a nasal cannula. I. Anesthesiology. 1988;68:659
6. Yanagidate F, Dohi S. Modified nasal cannula for simultaneous oxygen delivery and end-tidal CO2 monitoring during spontaneous breathing. Eur J Anaesthesiol. 2006;23:257–60
7. Roth JV, Barth LJ, Womack LH, Morgenlander LE. Evaluation of two commercially available carbon dioxide sampling nasal cannulae. J Clin Monit. 1994;10:237–43
8. Soto RG, Fu ES, Vila H Jr, Miguel RV. Capnography accurately detects apnea during monitored anesthesia care. Anesth Analg. 2004;99:379–82
9. Kasuya Y, Akça O, Sessler DI, Ozaki M, Komatsu R. Accuracy of postoperative end-tidal Pco2 measurements with mainstream and sidestream capnography in non-obese patients and in obese patients with and without obstructive sleep apnea. Anesthesiology. 2009;111:609–15
© 2015 International Anesthesia Research Society